CN102700572B - Method for monitoring and early warning visualization relay in block section for railway safe traveling - Google Patents

Abstract

The invention discloses a method for monitoring and early warning visualization relay in a block section for railway safe traveling. The method comprises the following specific steps: (1) carrying out system initialization; (2) sending video information in real time; (3) enabling a railway administration cloud monitoring center and a train-mounted monitoring system to store received monitoring videos respectively; (4) judging the railway occupation state with the video information; (5) judging the railway occupation state with audio information; (6) judging whether subgrade settlement occurs; (7) monitoring video relay; (8) monitoring train distance; and (9) carrying out warning. By the method, the problems that currently train drivers can not judge the distance to a train in front since the drivers can neither see nor hear the occupation condition of the railway within the safe distance in front are solved, and a support is provided for the safe traveling of the train and the prevention of accidents such as rear-end collision and derailment of the train under abnormal conditions such as railway occupation, slope landslide, subgrade settlement, bridge breakage and collapse, tunnel collapse, debris flow, dust storm, squall and torrential rain.

Description

Railway safe driving block section visual relay monitoring and early warning method

Technical field

The present invention relates to field of computer technology, further relate to railway safe driving block section visual relay monitoring and early warning method in vision computing technique and mobile communication technology field.Gps satellite position fixing system, GIS geographic information system, wire/wireless/communication technology of satellite, computer network, video monitoring and vision calculate and are organically blended in one by the present invention, visual control can be provided to railway safe driving block section, and safe driving is required that the front track in distance takies situation and carries out early warning, can as the important component part of intelligent traffic control system avoiding the accident such as train from overtaking collision, derailing.

Background technology

At present, usually ensure in railway operation that traffic safety uses block section autonomous cruise speed system, certain safe distance between vehicles is kept to allow between train, but once there is signal fault in this cover system, and because train operator cannot see the track occupation in front safe driving requirement distance clearly, just the grave accident such as train from overtaking collision, derailing may be caused.At present, for the monitoring of train traffic safety, existing following technology:

Patented technology " the railroad road safety safeguard system " (application number: 200820222334.3 that the Wei Chung people have, Authorization Notice No.: CN201291883Y) disclosed method is, driver wears and is used for observing when railway straight line is driven a vehicle infrared the looking in the distance of front road conditions to amplify TV anytime glasses, and arrange camera in railroad curve side, send video the track road conditions of telltale for driver observation curve ahead place to by wireless network.The deficiency that this patented technology exists is run at high speed in process at train, when driver finds that front is dangerous, when taking emergency braking measure, cannot meet the required distance of train safety braking.

The patent " the anti-way (necessary and sufficient condition) that knocks into the back of high speed railway " (application number: 201110256165.1, application publication number: CN102358329A) of David Wong application discloses a kind of technical scheme of high-speed rail rear-end collision prevention.The method is by continual direct contact between train department control personnel before and after high ferro, and the situations such as the instantaneous velocity of as prescribed circular train operation, direction and position, the abnormal condition between adjacent two trains of Timeliness coverage, to avoid generation car rear-end.The deficiency that the method exists is that the continual contact of manual type, is difficult to accomplish promptly and accurately.

Patented technology " the railway crossing comprehensive pre-warning device " (application number: 201120232912.3 that Jinan Tie Chengqi stone Electronics Co., Ltd. and Xiong Shiqi have, Authorization Notice No.: CN202124050U) disclosed method is, on-board screen is transferred to by the monitor video of the noctovisor by both sides, road junction, achieve when by the Special zone such as road junction, bend, train operator accurately can grasp the situation of forward box.The deficiency that this patented technology exists is, do not realize in train operation overall process, incessantly for train operator shows the situation of forward box, therefore, when abnormal condition appear in front track, because driver cannot Timeliness coverage and may cause the generation of accidents such as knocking into the back.

Patented technology " the automatic monitored control system for railway driving safety " (application number: 200720080920.4 that Siqi Communication Equipment Co., Ltd., Guilin City has, Authorization Notice No.: CN201214435Y) disclosed method is, by specifying unique section address code to CCTV camera, 3 ~ 6 kilometers, automatic monitoring front road conditions.The deficiency that this patented technology exists is, the distance in 3 ~ 6 kilometers, its automatic monitoring front can not meet the required distance of high speed train safety arrestment, and artificial to every platform CCTV camera assigned address code, be also unfavorable for the laying that CCTV camera is extensive, flexible, easy.

Summary of the invention

The present invention is directed to railway safe driving block section autonomous cruise speed system in above-mentioned prior art and may occur the deficiency of signal fault, and can not realize in prior art, to the over distance relay monitoring problem of 6 ~ 20 kilometers, proposing a kind of railway safe driving block section visual relay monitoring and early warning method relating to vision computing technique and mobile communication technology in the overall process of driving.

Realizing concrete thought of the present invention is, Intelligent monitoring camera relay is utilized to monitor the track of the multiple block section in front in train traveling process, allow train operator and road bureau's cloud monitoring and control centre can see the track occupation of 8 ~ 10 block sections in front by video-splicing technology, make train operator can hear the audio-frequency information of front track by Audio mixing, by each Intelligent monitoring camera real-time judge monitor taking situation and sending early warning information of track, the distance of the train in being advanced by GPS locating data real-time dynamic monitoring and its co-orbital front adjacent train, solve train operator in driving conditions and can not see and hear the track occupation of 8 ~ 10 block sections in front, the problem of the distance between the train of front can not be judged, thus for avoiding train from overtaking collision, the accident supplying methods such as derailing are supported.

The concrete steps that the present invention realizes are as follows:

(1) system initialization

1a) each Intelligent monitoring camera gathers the video information of track to be monitored respectively;

1b) track video information is sent to railway road bureau cloud monitoring and control centre by wireless channel;

1c) road bureau's cloud control center extracts the GPS locating data and altitude figures that comprise in track video information;

1d) store GPS locating data and altitude figures, as the initialization Reference data of the Intelligent monitoring camera of correspondence;

1e) vehicle-mounted monitoring system passes through the GPS Reference data of each Intelligent monitoring camera in section to road bureau's cloud monitoring and control centre request wish;

1f) utilize Canny edge detection operator to calculate the video image of each Intelligent monitoring camera, be stored as standard texture template image.

(2) video information is sent in real time

2a) each Intelligent monitoring camera gathers the video information of track to be monitored respectively;

2b) track video information is sent to road bureau's cloud monitoring and control centre and vehicle-mounted monitoring system by wireless channel.

(3) road bureau's cloud monitoring and control centre and vehicle-mounted monitoring system store received monitor video respectively.

(4) seizure condition of track is judged with video information

4a) RGB color image in each monitor video is converted into luminance picture by luminance picture conversion formula;

4b) luminance picture is carried out histogram equalization by histogram equalization formula;

4c) to histogram equalization, image carries out brightness normalization method by brightness normalization method formula again;

4d) utilize Canny edge detection operator to calculate normalized luminance picture, obtain its texture image;

4e) texture image and standard texture template image are subtracted each other, obtain error image;

4f) judge whether the energy of error image is greater than 10% of given standard texture template image energy, if be greater than, then think that track is occupied, perform step (9); Otherwise, perform step (2).

(5) seizure condition of track is judged with audio-frequency information

5a) the embedded sensor noise of Intelligent monitoring camera judges the seizure condition of track, if current noise decibel is greater than 60 decibels, then thinks that track is by the train occupation in advancing, and performs step (9); Otherwise, perform step (2);

5b) sound signal of each monitor video received is carried out audio mixing synthesis according to audio mixing composite formula by vehicle-mounted monitoring center, is play by acoustical equipment, manually monitors the seizure condition identifying front track for train operator.

(6) roadbed whether sedimentation is judged

Vehicle-mounted monitoring system extracts the GPS elevation information in each monitor video, with step 1d) in the Reference data of elevation subtract each other, judge whether the difference obtained is greater than given elevation threshold value, if be greater than, then think that monitored track exists sedimentation or caves in, perform step (9); Otherwise, perform step (2).

(7) video relay splicing

Vehicle-mounted monitoring system extracts the frame of same time point in each monitor video, according to the GPS locating data that each frame carries, by the frame that extracts with GPS locating data for sequence, adopt video relay joining method, relay is spliced into the real-time monitor video of monitored track, be presented on monitoring screen, for the seizure condition of train operator's manual observation front track.

(8) spacing is monitored

8a) according to the spacing between adjacent two trains each on spacing formulae discovery track, judge whether spacing is less than safety distance according to the speed of a motor vehicle, if be less than, perform step (9); Otherwise, perform step (2);

8b) road bureau's cloud monitoring and control centre is according to the GPS locating data of each monitor video, merges the dynamic monitoring video that generalized information system geography information generates whole train operation state on all circuits in administrative road bureau in real time, is presented in real time on monitoring screen.

(9) report to the police

9a) triggering voice warning function, prompting monitor staff process;

9b) trigger image alarm module, monitoring screen demarcates alert locations, does further process by monitor staff;

If 9c) monitor staff did not process in 10 seconds, system will trigger the automatic processing capacity of vehicle-mounted monitoring system alarm message;

9d) deal with emergencies and dangerous situations complete, go to step (2).

The present invention compared with prior art, has the following advantages:

First, because the present invention utilizes Intelligent monitoring camera relay to monitor the track occupation of the multiple block section in train direct of travel front, train operator can be allowed to see the track occupation of 8 ~ 10 block sections in front, solve the track occupation problem that driver in train traveling process can not see 8 ~ 10 block sections in front, the application of the method helps avoid the safety misadventure such as train from overtaking collision and derailing.

Second, because the present invention employs vision computing technique in video relay monitoring, what each Intelligent monitoring camera can judge institute's monitoring track takies situation, when abnormal condition appear in orbit occupancy, track slope collapse, subsidence, bridge, culvert, tunnel etc., send early warning information to avoid accident generations such as derailing and knock into the back.

3rd, because the present invention employs GPS locating data and generalized information system information in video relay monitoring, can real-time dynamic monitoring advance in train and the distance of its co-orbital front adjacent train, solve the distance problem that current train operator can not judge between the adjacent train of front, thus train departure density, the in advance accident such as avoid knocking into the back can be increased occur.

4th, owing to present invention uses the audio-frequency information in video relay monitoring, pass through Audio mixing, make train operator can hear the audio-frequency information of front track, assist train driver manually monitors the environmental information of front track and takies situation, thus can under the severe weather conditions such as strong wind, heavy rain, sandstorm, pre-decelerating and avoiding knocks into the back, the accident such as derailing occurs.

Accompanying drawing explanation

Fig. 1 is diagram of circuit of the present invention;

Fig. 2 is the keystone schematic diagram of step 7 video relay of the present invention splicing;

Fig. 3 is the former figure before keystone of the present invention;

Fig. 4 is the result figure after keystone of the present invention;

Fig. 5 is relay of the present invention splicing result figure.

Detailed description of the invention

Below in conjunction with accompanying drawing, the present invention is described in detail.

With reference to Fig. 1, realize concrete steps of the present invention as follows:

Step 1, system initialization

Set up high definition wireless/wired intelligent CCTV cameras by every block section along the line for monitored track at interval of 200 meters/400 meters positions, each Intelligent monitoring camera gathers track video information to be monitored respectively.Intelligent monitoring camera, comprises the detections of radar imaging system of ccd video camera, noctovisor and sub millimeter accuracy.By day or illumination condition is good when driven CCD camara module, and at night or thick fog, heavy rain or snowfall when drive the detections of radar imaging system of noctovisor or sub millimeter accuracy.

The video information of monitoring track is sent to railway road bureau cloud monitoring and control centre by wireless channel by each Intelligent monitoring camera, cloud control center of road bureau extracts the GPS locating data and altitude figures that comprise in track video information, store GPS locating data and altitude figures, as the initialization Reference data of the Intelligent monitoring camera of correspondence.Cloud control center of road bureau utilizes Canny edge detection operator to calculate the video image of each Intelligent monitoring camera, is stored as standard texture template image, as correspondence Intelligent monitoring camera judge the whether occupied standard texture template image of track;

Train-installed monitored control system to GPS Reference data for each Intelligent monitoring camera by section of road bureau cloud monitoring and control centre request, as the benchmark judging roadbed whether sedimentation.

Train-installed monitored control system passes through the standard texture template image of each Intelligent monitoring camera in section, as judging the whether occupied standard texture template image of track to road bureau's cloud monitoring and control centre request wish.

Step 2, sends video information in real time

Each Intelligent monitoring camera that roadbed sets up gathers track video information to be monitored respectively, vision computing technique is utilized whether to have abnormal object to appear at Abnormality Analysis on track, broadcast transmission track video information and alerting signal: " I is the monitoring camera on uplink/downlink line; I is * * * the current time; my GPS location is * * *; here discovery/and without noting abnormalities, my monitor video please be receive.”

Step 3, monitor video receives and stores

After vehicle-mounted monitoring system receives broadcast, judge whether oneself receives the video of this GPS location, if then receive and store transmitted monitor video, and to report to the police to train operator to carry out slowing down according to abnormal information, braking etc.; If not, then do nothing.

Road bureau's cloud monitoring and scheduling center stores corresponding monitor video after receiving each CCTV camera monitor video and alarm message, and carries out decision-making judgement, takes corresponding deceleration, brake measure to notify close to or to be in train in alarm range.

Step 4, judges the seizure condition of track with video information

The conversion formula that RGB color image in each monitor video is converted into luminance picture by following coloured image is converted into luminance picture I:

I＝(R+G+B)/3

Wherein, R, G, B represent the subimage that each component of red, green, blue in original color image is corresponding respectively.

The luminance picture I obtained is carried out histogram equalization by following luminance picture histogram equalization formula again, obtains the new images f of histogram equalization:

$S\left(k\right)=\left(\underset{j=0}{\overset{k}{\mathrm{\Σ}}}{n}_j\right)/N\×255$

Wherein, S (k) represents the gray value in original image after k gray level equalization, and k ∈ [0,255], Σ represents summation, n _jbe the pixel quantity of j gray level in original image, j ∈ [0, k], N are the sum of all pixels in original image.

The visibility that new images f adds than former luminance picture I and contrast ratio.The f of the image to histogram equalization is carried out brightness normalization method by following brightness normalization method formula again, obtains the difference of range of luminance values used when brightness normalization method image H obtains with removal of images:

$H(m,n)=\frac{f(m,n)-\min \left(f(m,n)\right)}{\max \left(f(m,n)\right)-\min \left(f(m,n)\right)}$

Wherein, H (m, n) ∈ [0,1] represents (m, n) the normalization method brightness value at pixel place, f represents pending image, and f (m, n) is for image f is at its pixel (m, n) brightness value at place, max (f (m, n)) and min (f (m, n)) represents maximum, the minimum luminance value of image f respectively.

Utilize Canny edge detection operator to calculate normalized luminance picture H, obtain its texture image E.Texture image E and standard texture template image are subtracted each other, obtains error image.Judge whether the energy of error image is greater than 10% of given standard texture template image energy, if be greater than, then think that track is occupied, perform step (9); Otherwise, perform step (2);

Step 5, judges the seizure condition of track with audio-frequency information:

The sound signal Xi (t) of each monitor video that vehicle-mounted monitoring center will receive, i=1 ~ 99, carry out audio mixing synthesis according to following audio mixing composite formula:

$m\left(t\right)=\frac{1}{n}\left(\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\mathrm{Xi}\left(t\right)\right),$

Wherein, m (t) represents the audio frequency after t audio mixing, and n represents and participates in the audio frequency number of audio mixing, Xi (t) be t i-th CCTV camera monitor the 8bit sampled value of audio frequency, the length of audio frame gets 10 ~ 20ms, and the sampling frequency of audio frame is identical.

Audio frequency m (t) after being synthesized by audio mixing is play by acoustical equipment, manually monitors the seizure condition identifying front track for train operator.Vehicle-mounted monitoring system judges the seizure condition of track according to the information of the embedded sensor noise of each Intelligent monitoring camera, if current noise decibel is greater than 60 decibels, then thinks that track is by the train occupation in advancing, and performs step (9); Otherwise, then step (2) is performed;

Step 6, judges roadbed whether sedimentation:

According to the height accuracy of current GPS, the scope of elevation threshold value is set to 3 ~ 5cm.Vehicle-mounted monitoring system extracts the GPS elevation information in each monitor video, subtract each other to the corresponding height datum data from road bureau's cloud monitoring and control centre request, judge whether the difference obtained is greater than given elevation threshold value, if be greater than, then think that monitored track exists sedimentation or caves in, perform step (9); Otherwise, perform step (2);

Step 7, video relay is spliced:

Vehicle-mounted monitoring system utilizes the video of continuous 1 ~ 99 CCTV camera in front and utilizes two-dimentional Mallat fast wavelet to carry out one deck wavelet decomposition respectively, extracts its low frequency component subimage.This low frequency component subimage is the compressed image of original image 1/4 size, maintains all detailed information and the visibility of original image.The low-frequency approximation component subimage interlacing obtained extracted, object is new two field picture I (k) carrying out recompressing to obtain reducing half, k=1 ~ 99 again.

In Fig. 2,4 bench mark (x are set ₁, y ₁), (x ₂, y ₂), (x ₃, y ₃), (x ₄, y ₄), then 4 and bench mark point (u are one to one set ₁, v ₁), (u ₂, v ₂), (u ₃, v ₃), (u ₄, v ₄).Point (x _i, y _i) and (u _i, v _i) mapping relations be

$\left\{\begin{array}{c}{u}_i=\frac{{\mathrm{\αx}}_i+{\mathrm{by}}_i+c}{{\mathrm{gx}}_i+{\mathrm{hy}}_i+1}\\ v_i=\frac{{\mathrm{dx}}_i+e{y}_i+f}{{\mathrm{gx}}_i+{\mathrm{hy}}_i+1}\end{array}\right.$

Wherein, (x _i, y _i) and (u _i, v _i) be the coordinate of respective pixel point in Fig. 2, a, b, c, d, e, f, g, h are mapping coefficients.

Separate following set of equations

$\left[\begin{array}{cccccccc}{x}_1& y_1& 1& 0& 0& 0& {-u}_1{x}_1& {-u}_1{y}_1\\ 0& 0& 0& x_1& y_1& 1& {-v}_1{x}_1& {-v}_1{y}_1\\ x_2& y_2& 1& 0& 0& 0& {-u}_2{x}_2& {-u}_2{y}_2\\ 0& 0& 0& x_2& y_2& 1& {-v}_2{x}_{\text{2}}& {-v}_2{y}_2\\ x_3& y_3& 1& 0& 0& 0& {-u}_3{x}_3& -u_3{y}_3\\ 0& 0& 0& x_3& y_3& 1& {-v}_3{x}_3& {-v}_3{y}_3\\ x_4& y_4& 1& 0& 0& 0& {-u}_4{x}_4& {-u}_4{y}_4\\ 0& 0& 0& x_4& y_4& 1& {-v}_4{x}_4& {-v}_4{y}_3\end{array}\right]\left[\begin{array}{c}a\\ b\\ c\\ d\\ e\\ f\\ g\\ h\end{array}\right]=\left[\begin{array}{c}{u}_1\\ v_1\\ u_2\\ v_2\\ u_3\\ v_3\\ u_4\\ v_4\end{array}\right],$

Obtain the Transformation Relation of Projection matrix between 4 bench marks and 4 corresponding point $\mathrm{cpt}=\left[\begin{array}{ccc}a& d& g\\ b& e& h\\ c& f& 1\end{array}\right]$

According to following keystone correction formula, correction process is carried out to image I (k), obtains image img (k) corrected.

$(x_i,y_i,1)\×\frac{1}{{\mathrm{\λ}}_i}\mathrm{cpu}=(u_i,v_i,1)$

Wherein, (x _i, y _i) be the coordinate of pixel i in image before keystone correction, λ _i=x _ig+y _ih+1, (u _i, v _i) be the coordinate of pixel i in new images img (k) after keystone correction

Adopt neighbour's method of interpolation to carry out interpolation to image img (k), obtain interpolation image A (k) that I (k) is corresponding.Be that sequence is spliced according to the following formula by each interpolation image A (k) by its raw GPS positioning data:

$\mathrm{cat}=\left[\begin{array}{c}{A}_1\\ A_2\\ \·\\ \·\\ \·\\ A_n\end{array}\right]$

Wherein, the new images that cat obtains after representing splicing, [] representing matrix, matrix A ₁, A ₂..., A _nrepresent the image participating in splicing respectively and there is identical line number, columns;

Every 1 second (or automatically setting interval time according to the speed of a motor vehicle), the stitching image cat that relay is monitored is presented on touch-screen display in real time, manually checks for train operator.

Step 8, monitoring spacing:

Road bureau's cloud monitoring and control centre, according to the GPS locating data of each monitor video, merges the dynamic monitoring video that generalized information system geography information generates whole train operation state on all circuits in administrative road bureau in real time, is presented in real time on monitoring screen; The GPS locating data sent in real time according to each train and the speed of a motor vehicle, calculate the distance between adjacent train, carry out safety distance monitoring and early warning.

Vehicle-mounted monitoring system calculates the spacing on same track between each adjacent two trains according to following spacing computing formula:

$d={[{(x_{i,t}-x_{j,t})}^2+{(y_{i,t}-y_{j,t})}^2]}^{\frac{1}{2}}$

Wherein, d represents the spacing between adjacent two trains; X represents longitude, and y represents latitude; I and j represents the numbering of train respectively, and t represents certain time point of extraction two train longitude and latitude data; (x _{i, t}, y _{i, t}), (x _{j, t}, y _{j, t}) represent the GPS longitude and latitude locating data that train i and j sends when moment t respectively.

If train gait of march is less than 120 kilometers/hour, safety distance is set to 4 ~ 6 block sections; If train gait of march is greater than 120 kilometers/hour, safety distance is set to 8 ~ 10 block sections.Judge whether gained spacing is less than safety distance according to the speed of a motor vehicle, if be less than, perform step (9); Otherwise, perform step (2);

Step 9, report to the police:

According to the noise threshold of sensor noise, triggering voice warning function, prompting monitor staff process.Or according to the judgement to image texture, trigger image alarm module, monitoring screen demarcates alert locations, does further process by monitor staff.If after language alarm or image alarm module start 10 seconds, monitor staff does not process in time, system, by triggering the automatic processing capacity of vehicle-mounted monitoring system alarm message, turns the autonomous cruise speed system of train.After dealing with emergencies and dangerous situations, go to step (2), then reenter video relay monitor procedure.

Video relay splicing effect of the present invention further illustrates by following emulation experiment.

1. emulation experiment condition

This emulation utilizes MATLAB software to emulate.

2. emulate content

Emulation experiment of the present invention mainly carries out the emulation of keystone correction and relay splicing to monitor video two field picture.Fig. 3 uses the width size that obtained by video monitoring camera of prior art to be the orbit monitoring image of 400 × 200 pixels; Fig. 4 is the distortion using prior art to occur to Fig. 3, and the result figure obtained after adopting keystone distortion correction method of the present invention to carry out keystone correction; Fig. 5 be to Fig. 4 keystone correction after image, the analogous diagram obtained after using relay joining method of the present invention to carry out relay splicing; In Fig. 5, each keystone image participating in splicing, is that the GPS locating data corresponding to the orbit monitoring image of its correspondence is that sequence carries out splicing.

3. simulation result

As can be seen from the orbital image in Fig. 3, owing to using decorating position and its convex lens focus image-forming principle of the CCTV camera of prior art, make two railway lines be originally parallel to each other in obtained image lose parallel relation, occur the keystone distortion that near-end is large and far-end is little.As can be seen from the orbital image in Fig. 4, after using keystone correction of the present invention, uneven two railway lines in Fig. 3 are made to have recovered its due parallel relation originally.As can be seen from the relay splicing simulation result in Fig. 5, according to method of the present invention, the monitor video of continuous for train front 30 ~ 100 CCTV cameras can be carried out keystone correction, carry out relay splicing again, realize the video monitor shown on a frame 6 ~ 20 kilometers of track road conditions in front, make train operator over the horizon can see the situations such as the orbit occupancy in front like a cork.

Claims (10)

1. a railway safe driving block section visual relay monitoring and early warning method, comprises the steps:

(1) system initialization

1a) each Intelligent monitoring camera gathers the video information of track to be monitored respectively;

1b) track video information is sent to railway road bureau cloud monitoring and control centre by wireless channel;

1c) railway road bureau cloud monitoring and control centre extracts the GPS locating data and altitude figures that comprise in track video information;

1d) store GPS locating data and altitude figures, as the initialization Reference data of the Intelligent monitoring camera of correspondence;

1e) vehicle-mounted monitoring system passes through the GPS Reference data of each Intelligent monitoring camera in section to railway road bureau cloud monitoring and control centre request wish;

1f) utilize Canny edge detection operator to calculate the video image of each Intelligent monitoring camera, be stored as standard texture template image;

(2) video information is sent in real time

2a) each Intelligent monitoring camera gathers the video information of track to be monitored respectively;

2b) track video information is sent to railway road bureau cloud monitoring and control centre and vehicle-mounted monitoring system by wireless channel;

(3) railway road bureau cloud monitoring and control centre and vehicle-mounted monitoring system store received monitoring video information respectively;

(4) seizure condition of track is judged with video information

4a) RGB color image in each monitoring video information is converted into luminance picture by luminance picture conversion formula;

4b) luminance picture is carried out histogram equalization by histogram equalization formula;

4c) by brightness normalization method formula, brightness normalization method is carried out to the image of histogram equalization;

4d) utilize Canny edge detection operator to calculate normalized luminance picture, obtain its texture image;

4e) texture image and standard texture template image are subtracted each other, obtain error image;

4f) judge whether the energy of error image is greater than 10% of given standard texture template image energy, if be greater than, then think that track is occupied, perform step (9); Otherwise, perform step (2);

(5) seizure condition of track is judged with audio-frequency information

5a) the embedded sensor noise of Intelligent monitoring camera judges the seizure condition of track, if current noise decibel is greater than 60 decibels, then thinks that track is by the train occupation in advancing, and performs step (9); Otherwise, perform step (2);

5b) sound signal of each monitoring video information received is carried out audio mixing synthesis according to audio mixing composite formula by vehicle-mounted monitoring center, is play by acoustical equipment, manually monitors the seizure condition identifying front track for train operator;

(6) roadbed whether sedimentation is judged

Vehicle-mounted monitoring system extracts the GPS elevation information in each monitoring video information, with step 1d) in the Reference data of elevation subtract each other, judge whether the difference obtained is greater than given elevation threshold value, if be greater than, then think that monitored track exists sedimentation or caves in, perform step (9); Otherwise, perform step (2);

(7) video relay splicing

Vehicle-mounted monitoring system extracts the frame of same time point in each monitoring video information, according to the GPS locating data that each frame carries, by the frame that extracts with GPS locating data for sequence, adopt video relay joining method, relay is spliced into the real-time monitoring video information of monitored track, be presented on monitoring screen, for the seizure condition of train operator's manual observation front track;

(8) spacing is monitored

8a) according to the spacing between adjacent two trains each on spacing formulae discovery track, judge whether spacing is less than safety distance according to the speed of a motor vehicle, if be less than, perform step (9); Otherwise, perform step (2);

8b) railway road bureau cloud monitoring and control centre is according to the GPS locating data of each monitoring video information, merge the dynamic monitoring video information that generalized information system geography information generates whole train operation state on all circuits in administrative road bureau in real time, be presented in real time on monitoring screen;

(9) report to the police

9a) triggering voice warning function, prompting monitor staff process;

9b) trigger image alarm module, monitoring screen demarcates alert locations, does further process by monitor staff;

If 9c) monitor staff did not process in 10 seconds, system will trigger the automatic processing capacity of vehicle-mounted monitoring system alarm message;

9d) deal with emergencies and dangerous situations complete, go to step (2).

2. railway safe driving block section according to claim 1 visual relay monitoring and early warning method, it is characterized in that: step 1a) described in each Intelligent monitoring camera, be at interval of 200 meters/400 meters positions erections in track block section along the line every.

3. railway safe driving block section according to claim 1 visual relay monitoring and early warning method, is characterized in that: step 4a) described in coloured image to be converted into the conversion formula of luminance picture as follows:

I＝(R+G+B)/3

Wherein, I represents luminance picture, and R, G, B represent the subimage that each component of the red, green, blue of coloured image is corresponding respectively.

4. railway safe driving block section according to claim 1 visual relay monitoring and early warning method, is characterized in that: step 4b) described in luminance picture histogram equalization formula as follows:

$S\left(k\right)=\left(\underset{j=0}{\overset{k}{\mathrm{\Σ}}}{n}_j\right)/N\×255$

Wherein, S (k) represents the gray value in original image after k gray level equalization, and k ∈ [0,255], ∑ represents summation, n _jbe the pixel quantity of j gray level in original image, j ∈ [0, k], N are the sum of all pixels in original image.

5. railway safe driving block section according to claim 1 visual relay monitoring and early warning method, is characterized in that: step 4c) described in brightness normalization method formula as follows:

$H(m,n)=\frac{f(m,n)-\min \left(f(m,n)\right)}{\max \left(f(m,n)\right)-\min \left(f(m,n)\right)}$

Wherein, H (m, n) ∈ [0,1] represents (m, n) the normalization method brightness value at pixel place, f represents pending image, and f (m, n) is for image f is at its pixel (m, n) brightness value at place, max (f (m, n)) and min (f (m, n)) represents maximum, the minimum luminance value of image f respectively.

6. railway safe driving block section according to claim 1 visual relay monitoring and early warning method, is characterized in that: step 5b) described in audio mixing composite formula as follows:

$m\left(t\right)=\frac{1}{n}\left(\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\mathrm{Xi}\left(t\right)\right)$

Wherein, m (t) represents the audio frequency after t audio mixing, and n represents and participates in the audio frequency number of audio mixing, Xi (t) be t i-th CCTV camera monitor the 8bit sampled value of audio frequency, the length of audio frame gets 10 ~ 20ms, and the sampling frequency of audio frame is identical.

7. railway safe driving block section according to claim 1 visual relay monitoring and early warning method, is characterized in that: the scope of the elevation threshold value described in step (6) is 3 ~ 5cm.

8. railway safe driving block section according to claim 1 visual relay monitoring and early warning method, is characterized in that: the concrete steps of the video relay joining method described in step (7) are as follows:

The first, utilize two-dimentional Mallat fast wavelet to carry out one deck wavelet decomposition respectively to the video of the continuous multiple CCTV camera in front, extract its low frequency component subimage;

The second, by the low-frequency approximation component subimage interlacing extraction again obtained, obtain the new two field picture reducing half;

3rd, new two field picture is carried out keystone correction again;

4th, the two field picture by keystone is that sequence is spliced according to the following formula by its raw GPS positioning data:

$\mathrm{cat}=\left[\begin{array}{c}{A}_1\\ A_2\\ .\\ .\\ .\\ A_n\end{array}\right]$

Wherein, the new images that cat obtains after representing splicing, [] representing matrix, matrix A ₁, A ₂..., A _nrepresent the image participating in splicing respectively and there is identical line number, columns.

9. railway safe driving block section according to claim 1 visual relay monitoring and early warning method, is characterized in that: step 8a) described in spacing formula as follows:

$d={[{(x_{i,t}-x_{j,t})}^2+{(y_{i,t}-y_{j,t})}^2]}^{\frac{1}{2}}$

Wherein, d represents the spacing between adjacent two trains; X represents longitude, and y represents latitude; I and j represents the numbering of train respectively, and t represents certain time point of extraction two train longitude and latitude data; (x _{i, t}, y _{i, t}), (x _{j, t}, y _{j, t}) represent the GPS longitude and latitude locating data that train i and j sends when moment t respectively.

10. railway safe driving block section according to claim 1 visual relay monitoring and early warning method, is characterized in that: step 8a) described in adjacent train between safety distance, adopt with the following method arrange:

If train gait of march is less than 120 kilometers/hour, safety distance is set to 4 ~ 6 block sections; If train gait of march is greater than 120 kilometers/hour, safety distance is set to 8 ~ 10 block sections.